A test system and method for simulating speech recognition success rate at different distances

Abstract

The present invention discloses a test system and method for simulating speech recognition success rates at different distances. The test system for simulating speech recognition success rates at different distances includes: a calibration module for calibrating speech recognition success rates at different distances with a fixed sound source loudness; a test module for obtaining different loudnesses of a sound source at a fixed distance according to the speech recognition success rates at different distances; and a simulation module for simulating speech recognition success rates of a fixed sound source loudness at different distances according to the obtained different loudnesses of a sound source at a fixed distance. The present invention changes the loudness of a sound source at a fixed position to simulate speech recognition success rates at different distances, thereby solving the problem of space limitations in the testing process in the prior art.

Description

Technical Field

[0001] The present invention belongs to the technical field of speech recognition, and in particular relates to a testing system and method for simulating speech recognition success rates at different distances. Background Art

[0002] Speech recognition success rate is a key metric for speech recognition products (such as smart speakers). Speech recognition success rates vary at different distances. Generally, the greater the distance between the sound source and the device under test, the lower the speech recognition success rate; conversely, the higher the distance. During product testing and evaluation, it's important to measure the product's maximum recognition distance while maintaining a reasonable recognition success rate, thereby assessing its performance.

[0003] Traditional testing method: fix the volume of the sound source, adjust the distance between the device under test and the sound source, and test the recognition success rate at different distances.

[0004] Disadvantages of traditional testing methods:

[0005] 1. The existing method requires constant adjustment of the distance between the device under test and the sound source during the testing process, and each test takes a long time;

[0006] 2. The test method is limited by space. If the laboratory or test environment is small, it may not be possible to test the required data.

[0007] As disclosed in the invention patent with publication number CN106548772A, the speech recognition test system and method include: an audio generation module for generating an audio file from a test text based on test parameters; a speech output module for playing the audio file to provide speech input to the speech recognizer to be tested; and an information processing module for processing the speech recognition results from the speech recognizer to obtain a test report about the speech recognizer, the test report including recognition performance parameters of the speech recognizer under different conditions associated with the test parameters.

[0008] This invention patent can analyze the maximum distance between the identifier and the user under the premise that the identifier can work. If the test environment is limited, it is impossible to accurately determine the maximum distance between the identifier and the user under the premise that the identifier can work.

[0009] Therefore, the purpose of the present invention is to provide a method for testing speech recognition success rates at simulated distances, which is not limited by space.

[0010] The above content is only used to assist in understanding the technical solution of the present invention and does not constitute an admission that the above content is prior art. Summary of the Invention

[0011] In response to the above technical problems existing in the prior art, the present invention provides a testing system and method for simulating the success rate of speech recognition at different distances, which simulates the success rate of speech recognition at different distances from a fixed sound source by adjusting the volume of different sound sources at a fixed distance.

[0012] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0013] A test system for simulating speech recognition success rates at different distances, comprising:

[0014] A calibration module, used to calibrate the speech recognition success rate at different distances with a fixed source loudness;

[0015] A testing module, configured to obtain different loudnesses of a sound source at a fixed distance according to the speech recognition success rate obtained by the calibration module;

[0016] The simulation module is used to simulate the speech recognition success rate of the fixed sound source at different distances according to the different loudness of the sound source at the fixed distance.

[0017] As a preferred embodiment of the present invention, the calibration module includes:

[0018] An acquisition unit, used to set the loudness and position of the pronunciation source and obtain the corresponding speech recognition success rate;

[0019] The position moving unit is used to change the position of the pronunciation source and record the corresponding speech recognition success rate after the pronunciation source position is changed.

[0020] As a preferred embodiment of the present invention, the test module includes:

[0021] An adjustment unit is used to test the loudness of a corresponding pronunciation source at a fixed distance to ensure that the speech recognition success rate is consistent with that obtained by the calibration module;

[0022] The traversal unit is used to traverse the different speech recognition success rates obtained by the calibration module and record the loudness of the pronunciation source corresponding to the fixed distance.

[0023] As a preferred embodiment of the present invention, the simulation module includes:

[0024] The acquisition unit simulates the success rate of speech recognition of fixed sound source loudness at different distances according to the fixed sound source loudness.

[0025] The result statistics unit changes the loudness of the sound source at the fixed distance and calculates the speech recognition success rates corresponding to the different loudnesses of the sound source at the fixed distance.

[0026] As a preferred embodiment of the present invention, the test module further comprises: changing different fixed distances, and testing different loudnesses of the pronunciation source corresponding to the changed fixed distances according to the speech recognition success rate obtained by the calibration module.

[0027] A method for testing speech recognition success rates at simulated distances, comprising:

[0028] S1, calibrating the speech recognition success rate at different distances with a fixed source loudness;

[0029] S2. obtaining different loudnesses of a pronunciation source at a fixed distance according to the speech recognition success rates at different distances;

[0030] S3. According to the obtained different loudnesses of the sound source at a fixed distance, the speech recognition success rates of the fixed sound source loudness at different distances are simulated.

[0031] As a preferred embodiment of the present invention, step S1 includes:

[0032] S11. Set the loudness and location of the pronunciation source and obtain the corresponding speech recognition success rate;

[0033] S12. Change the pronunciation source position and record the corresponding speech recognition success rate after the pronunciation source position is changed.

[0034] As a preferred embodiment of the present invention, step S2 includes:

[0035] S21. At a fixed distance, ensuring that the speech recognition success rate is consistent with that obtained in step S1, test the loudness of the corresponding pronunciation source;

[0036] S22 , traverse the different speech recognition success rates obtained in step S1 , and record the loudness of the pronunciation source corresponding to the fixed distance.

[0037] As a preferred embodiment of the present invention, step S3 includes:

[0038] S31, using fixed-distance sound source loudness at different distances corresponding to the fixed-distance sound source loudness, simulating the speech recognition success rate of fixed-distance sound source loudness at different distances;

[0039] S32: Changing the loudness of the sound source at the fixed distance, and calculating the speech recognition success rates corresponding to different loudnesses of the sound source at the fixed distance.

[0040] As a preferred embodiment of the present invention, the step S2 further includes: S200, changing different fixed distances, and testing different loudnesses of the pronunciation sources corresponding to the changed fixed distances according to the speech recognition success rate obtained in the step S1.

[0041] The technical solution provided by the present invention can have the following beneficial effects:

[0042] 1. The present invention obtains a relationship diagram between distance and sound source volume through pre-calibration. This relationship diagram is used to test the sound source volume required at fixed distances for different speech recognition success rates. In subsequent speech recognition success rate tests, the sound source volume is varied at a fixed distance to simulate speech recognition success rates at different distances from a fixed sound source, simplifying the method for testing recognition success rates at different distances.

[0043] 2. The calibration module and the test module in the scheme of the present invention are implemented as a one-time behavior. After the calibration and testing of the standard data are successful, the sound source size can be set in the actual test in the future to simulate different distance conditions. The test scheme of the present invention reduces the test cost and simplifies the test process. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is an example diagram of the calibration process of the calibration module in embodiment 1 of the present invention;

[0045] Figure 2 A statistical table of the results obtained by the calibration module in Example 1 of the present invention;

[0046] Figure 3 This is a line chart of the results obtained by the calibration module in Example 1 of the present invention;

[0047] Figure 4 This is a statistical table of the test results obtained by the test module in Example 1 of the present invention;

[0048] Figure 5 This is a structural diagram of a test system for simulating speech recognition success rates at different distances according to Example 1 of the present invention;

[0049] Figure 6 This is a flow chart of a testing method for simulating speech recognition success rates at different distances according to Example 2 of the present invention. DETAILED DESCRIPTION

[0050] The present invention will now be more fully described with reference to the accompanying drawings included hereinafter, in which embodiments of the present invention are shown. However, the present invention may be embodied in many different forms and should not be construed as limited to the specific embodiments set forth herein. Rather, these embodiments are intended to convey the scope of the present invention to those skilled in the art.

[0051] Unless otherwise defined, the terms used herein (including technical and scientific terms) should be understood to have the same meaning as commonly understood by those skilled in the art to which the present invention pertains. Furthermore, it should be understood that the terms used herein should be understood to have the same meaning as that in this specification and the relevant art, and should not be interpreted in an ideal or overly formal sense unless expressly provided otherwise herein.

[0052] Example 1

[0053] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0054] This embodiment provides a test system for simulating the success rate of speech recognition at different distances, such as Figure 5 As shown, it includes: a calibration module 100, an acquisition unit 110, a position movement unit 120, a test module 200, an adjustment unit 210, a traversal unit 220, a simulation module 300, a collection unit 310, and a result statistics unit 320. The specific operation process of this solution is as follows:

[0055] The calibration module 100 is used to calibrate the speech recognition success rate at different distances with a fixed sound source loudness.

[0056] The calibration module 100 is used to calibrate the speech recognition success rate of the device under test when the pronunciation source is set at different distances from the device under test under normal circumstances.

[0057] The calibration module 100 includes an acquisition unit 110 and a position movement unit 120 . The calibration module 100 implements the above solution through the acquisition unit 110 and the position movement unit 120 .

[0058] The acquisition unit 110 is used to set the loudness and position of the pronunciation source and obtain the corresponding speech recognition success rate.

[0059] like Figure 1 As shown in the example diagram, the device under test is a calibration MIC (microphone) for receiving voice signals.

[0060] For example, the loudness of the sound source is set to 60dB, which is the sound of normal speech; and the position of the sound source is set to be 0.5m away from the device under test. The test of this embodiment is carried out in a quiet indoor environment (30dB).

[0061] The speech recognition success rate of the calibrated MIC is 55%.

[0062] The position moving unit 120 is used to change the position of the pronunciation source and record the corresponding speech recognition success rate after the pronunciation source position is changed.

[0063] The position moving unit 120 moves the position of the sound source. This embodiment simulates the distance of the living room in an ordinary home scene, so that the sound source varies between 0.5m and 7m, with a distance interval of 0.5m. Starting from a distance of 0.5m, the sound source is collected every 0.5m, and ends at 7m.

[0064] Each time the position of the pronunciation source is moved, the pronunciation source pronounces at a fixed volume at that position, and the success rate of speech recognition received by the MIC is calibrated at different distances.

[0065] The volume level received by the MIC is the success rate of speech recognition. Figure 2 Statistics table, Figure 3 As shown in the broken line statistical graph, the volume levels received by the calibrated MIC corresponding to distances from 0.5m to 7m are obtained.

[0066] The testing module 200 is configured to test different loudnesses of a sound source at a fixed distance according to the speech recognition success rates at different distances.

[0067] The testing module 200 measures different loudnesses of a sound source at a fixed distance based on the speech recognition success rate obtained by the calibration module 100 .

[0068] The testing module 200 further includes an adjusting unit 210 and a traversing unit 220 .

[0069] The adjustment unit is used to test the loudness of the corresponding pronunciation source at a fixed distance to ensure that the speech recognition success rate is consistent with that obtained by the calibration module 100.

[0070] Based on the recognition success rate obtained in the calibration module 100 , the sound source loudness is adjusted at a fixed position to obtain a sound source loudness corresponding to the recognition success rate obtained in the calibration module 100 .

[0071] For example, the fixed distance is set to 0.5 m, and the recognition success rate obtained in the calibration module 100 is 52, so that the volume obtained by calibrating the MIC is 52, and the size of the pronunciation source is adjusted to obtain a volume of 57.

[0072] The traversal unit 220 is configured to traverse the different speech recognition success rates obtained by the calibration module 100 and record the loudness of the sound source corresponding to a fixed distance.

[0073] The traversal unit 220 ensures that the volume obtained by calibrating the MIC is consistent with the speech recognition success rate obtained by the calibration module 100, and records the loudness of the sound source corresponding to the fixed distance.

[0074] For example, if the fixed distance is set to 0.5m, and the speech recognition success rates obtained by traversing the calibration module 100 are 55, 52, 50, ..., 38, 37, and the volume obtained by calibrating the MIC is consistent with 55, 52, 50, ..., 38, 37, the required sound source loudness is recorded; the sound source loudness corresponding to 55, 52, 50, ..., 38, 37 is 60, 57, 55, ..., 43, 42, respectively. Figure 4 As shown in the statistical table, the data in the table are schematic data and may deviate from the actual data.

[0075] The above describes the test distance calibration process. This involves pre-researching and testing the distance between the sound source and the device under test, while maintaining the same sound source volume. The two-dimensional relationship between the sound source volume received by the device under test and the distance to the sound source is recorded. The distance to the sound source is then fixed, and the sound source volume is adjusted to simulate changes in the source's distance. The specific standard for adjusting the sound source volume is based on the speech recognition success rate obtained by the calibration module.

[0076] Preferably, the testing module 200 further includes: changing different fixed distances, and testing different loudnesses of the pronunciation source corresponding to the changed fixed distances according to the speech recognition success rate obtained by the calibration module 100.

[0077] For example, changing the fixed distance to 1 meter ensures that the volume obtained by the calibration MIC is consistent with the volume in the calibration module 100, and the corresponding sound source loudness is obtained. The specific operation method is consistent with the operation method of the test module 200 described above and will not be repeated here. Changing the fixed distance allows users to perform simulation tests at another fixed distance in the future when it is inconvenient to perform simulation tests at a fixed location.

[0078] The simulation module 300 is used to simulate the speech recognition success rate of the fixed sound source at different distances according to the obtained different loudness of the sound source at the fixed distance.

[0079] Set the volume of the pronunciation source as required at a fixed location to simulate the success rate of speech recognition at different distances.

[0080] The simulation module 300 is composed of a collection unit 310 and a result statistics unit 320 to complete the simulation process. The implementation scheme is as follows:

[0081] The acquisition unit 310 simulates the speech recognition success rate of the fixed-distance sound source loudness at different distances corresponding to the fixed-distance sound source loudness;

[0082] For example: the volume obtained by the test module 200 when the fixed distance is 0.5m and the loudness of the sound source is 42 is consistent with the volume obtained by calibrating the MIC when the distance is 7m and the loudness of the fixed sound source is 60. Therefore, during later maintenance or testing, the fixed distance of 0.5m and the loudness of the sound source can be used as parameters to simulate the volume obtained when the distance is 7m and the loudness of the fixed sound source is 60.

[0083] The result statistics unit 320 changes the loudness of the sound source at the fixed distance, and calculates the speech recognition success rates corresponding to the different loudness of the sound source at the fixed distance.

[0084] For example, at a location 0.5 m away from the calibration MIC, the loudness of the sound source is changed from 60 to 42, the method of the acquisition unit 310 is repeated, and the speech recognition success rates obtained are counted.

[0085] The speech recognition success rate obtained by the result statistics unit 320 is the speech recognition success rate obtained by fixing the sound source volume and adjusting the sound source position.

[0086] To sum up, this embodiment provides a test system for simulating the success rate of speech recognition at different distances. The calibration module in the scheme and the scheme in the test module are one-time behaviors when implemented. After the standard data is successfully calibrated and tested, different distances can be simulated in subsequent actual tests by setting the sound source size. The test scheme of the present invention reduces the test cost and simplifies the test process.

[0087] Example 2

[0088] This embodiment provides a method for testing the success rate of speech recognition at different distances. Figure 6 As shown, the specific process may include the following steps:

[0089] S1. Calibrate the speech recognition success rate at different distances with a fixed source loudness.

[0090] Calibration is generally performed to determine the speech recognition success rate of the device under test when the sound source is at different distances from the device under test.

[0091] The step S1 comprises:

[0092] S11. Set the loudness and position of the pronunciation source to obtain the corresponding speech recognition success rate.

[0093] The device under test is a calibration MIC (microphone), which is used to receive voice signals. The volume received by the calibration MIC is the voice recognition success rate.

[0094] The sound source is a device that can make a sound, such as a speaker, a mobile phone, etc.

[0095] For example, the loudness of the sound source is set to 60dB, which is the sound of normal speech; and the position of the sound source is set to be 0.5m away from the device under test. The test of this embodiment is carried out in a quiet indoor environment (30dB).

[0096] The speech recognition success rate of the calibrated MIC is 55%.

[0097] S12. Change the pronunciation source position and record the corresponding speech recognition success rate after the pronunciation source position is changed.

[0098] The position of the sound source is moved. This embodiment simulates the distance of the living room in an ordinary home scene, and the sound source is changed between 0.5m and 7m, with a distance interval of 0.5m. Starting from a distance of 0.5m, the sound source is collected every 0.5m, and ends at 7m.

[0099] Each time the sound source is moved, it speaks at a fixed volume at that location, and the speech recognition success rate received by the calibrated MIC at different distances is recorded. The speech recognition success rates corresponding to distances from 0.5m to 7m are obtained.

[0100] S2. Obtain different loudnesses of a pronunciation source at a fixed distance according to the speech recognition success rates at different distances.

[0101] In this step, the different loudnesses of the sound source at a fixed distance are measured based on the speech recognition success rate obtained in step S1.

[0102] The step S2 comprises:

[0103] S21 . At a fixed distance, ensuring that the speech recognition success rate is consistent with that obtained in step S1 , the loudness of the corresponding pronunciation source is tested.

[0104] Based on the recognition success rate obtained in step S1, the sound source loudness is adjusted at a fixed position to obtain the sound source loudness corresponding to the recognition success rate obtained in step S1.

[0105] For example, the fixed distance is set to 0.5 m, and the recognition success rate obtained in step S1 is 52, so that the volume obtained by calibrating the MIC is 52, and the size of the pronunciation source is adjusted to obtain a volume of 57.

[0106] S22 , traverse the different speech recognition success rates obtained in step S1 , and record the loudness of the pronunciation source corresponding to the fixed distance.

[0107] Ensure that the volume obtained by calibrating the MIC is consistent with the speech recognition success rate obtained in step S1, and record the loudness of the sound source corresponding to the fixed distance.

[0108] For example, a fixed distance of 0.5 m is set, and the speech recognition success rates obtained in step S1 are 55, 52, 50, ..., 38, and 37, respectively. The required sound source loudness is ensured when the volume obtained by calibrating the MIC is consistent with 55, 52, 50, ..., 38, and 37, respectively; the sound source loudness corresponding to 55, 52, 50, ..., 38, and 37 is recorded as 60, 57, 55, ..., 43, and 42, respectively.

[0109] The step S2 further includes: S200, changing different fixed distances, and testing different loudnesses of the pronunciation sources corresponding to the changed fixed distances according to the speech recognition success rate obtained in the step S1.

[0110] For example, changing the fixed distance to 1m ensures that the volume obtained by calibrating the MIC is consistent with the volume in step S1, and the corresponding sound source loudness is obtained. The specific operation method is the same as the operation method in step S2 above and is not repeated here. Changing the fixed distance allows users to simulate at another fixed distance when it is inconvenient to perform simulation tests at a fixed location.

[0111] S3. According to the obtained different loudnesses of the sound source at a fixed distance, the speech recognition success rates of the fixed sound source loudness at different distances are simulated.

[0112] Set the volume of the pronunciation source as required at a fixed location to simulate the success rate of speech recognition at different distances.

[0113] The step S3 comprises:

[0114] S31. Using the fixed-distance sound source loudness at different distances corresponding to the fixed-distance sound source loudness, simulate the speech recognition success rate of the fixed-distance sound source loudness at different distances.

[0115] For example: the volume obtained by step S2 when the fixed distance is 0.5m and the loudness of the sound source is 42 is consistent with the volume obtained by calibrating the MIC when the distance is 7m and the loudness of the fixed sound source is 60. Therefore, during later maintenance or testing, the fixed distance of 0.5m and the loudness of the sound source can be used as parameters to simulate the volume obtained when the distance is 7m and the loudness of the fixed sound source is 60.

[0116] S32: Changing the loudness of the sound source at the fixed distance, and calculating the speech recognition success rates corresponding to different loudnesses of the sound source at the fixed distance.

[0117] For example, at a location 0.5 m away from the calibration MIC, the loudness of the sound source is changed from 60 to 42, and the loudness of the sound source corresponding to all distances is traversed. The method of step S31 is repeated to calculate the speech recognition success rates obtained respectively.

[0118] The speech recognition success rate obtained in the above step S32 is the speech recognition success rate obtained by fixing the sound source volume and adjusting the sound source position.

[0119] In summary, this embodiment provides a test method for simulating the success rate of speech recognition at different distances. First, the sound source size is fixed, and the volume level received by the calibration MIC is recorded at different distances to obtain the relationship between the distance and the volume level received by the calibration MIC. Then, the volume level received by the MIC is ensured to be consistent with the volume level received by the MIC. At a fixed distance, the sound source volume level is adjusted to obtain the corresponding sound source volume level. Subsequently, the speech recognition success rate at different distances with a fixed sound source size is simulated using different sound source volume levels at a fixed distance.

[0120] The implementation of the scheme in step S1 and step S2 in this embodiment is a one-time behavior. After the standard data is successfully calibrated and tested, different distance conditions can be simulated in subsequent actual tests by setting the sound source size. The test scheme of the present invention reduces test costs and simplifies the test process.

[0121] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the following claims.

Claims

1. A test system for simulating speech recognition success rates at different distances, characterized in that: include: A calibration module, used to calibrate the success rate of speech recognition at different distances with a fixed sound source loudness; A testing module, used to obtain different loudnesses of a sound source at a fixed distance according to the speech recognition success rate obtained by the calibration module; A simulation module, used for simulating the speech recognition success rate of the fixed sound source loudness at different distances according to the different loudness of the sound source at the fixed distance; The test module includes: An adjustment unit is used to test the loudness of the corresponding sound source at a fixed distance to ensure that the speech recognition success rate is consistent with that obtained by the calibration module; A traversal unit, used for traversing different speech recognition success rates obtained by the calibration module, and recording the loudness of the sound source corresponding to the fixed distance; The simulation module includes: a collection unit, which simulates the speech recognition success rate of the fixed sound source loudness at different distances according to the fixed sound source loudness at different distances; and a result statistics unit, which changes the sound source loudness at the fixed distance and counts the speech recognition success rate corresponding to the different loudness of the sound source at the fixed distance.

2. The test system for simulating speech recognition success rates at different distances according to claim 1, characterized in that: The calibration module comprises: An acquisition unit, used to set the loudness and position of a pronunciation source and acquire a corresponding speech recognition success rate; The position moving unit is used to change the position of the pronunciation source and record the corresponding speech recognition success rate after the position of the pronunciation source is changed.

3. The test system for simulating speech recognition success rates at different distances according to claim 1, characterized in that: The testing module further includes: changing different fixed distances, and testing different loudnesses of the pronunciation sources corresponding to the changed fixed distances according to the speech recognition success rate obtained by the calibration module.

4. A method for testing the success rate of speech recognition at different distances, characterized in that: include: S1, calibrate the speech recognition success rate at different distances with a fixed source loudness; S2. obtaining different loudnesses of the pronunciation source at a fixed distance according to the speech recognition success rates at different distances; S3, according to the different loudness of the sound source at a fixed distance, the corresponding speech recognition success rate of the fixed sound source loudness at different distances is simulated; The step S2 comprises: S21, at a fixed distance, ensuring that the speech recognition success rate is consistent with that obtained in the step S1, testing the loudness of the corresponding sound source; S22, traversing the different speech recognition success rates obtained in the step S1, and recording the loudness of the sound source corresponding to the fixed distance; The step S3 comprises: S31, simulating the speech recognition success rate of the fixed sound source loudness at different distances corresponding to the fixed sound source loudness; S32, changing the sound source loudness at the fixed distance, and counting the speech recognition success rate corresponding to the different loudness of the sound source at the fixed distance.

5. The method for testing the success rate of speech recognition at different distances according to claim 4, characterized in that: The step S1 comprises: S11, setting the loudness and position of the pronunciation source, and obtaining the corresponding speech recognition success rate; S12. Change the pronunciation source position, and record the corresponding speech recognition success rate after the pronunciation source position is changed.

6. The method for testing the success rate of speech recognition at different distances according to claim 4, characterized in that: The step S2 further includes: S200, changing different fixed distances, and testing different loudnesses of the pronunciation sources corresponding to the changed fixed distances according to the speech recognition success rate obtained in the step S1.

Images